On a a posteriori error estimator for the discontinuous Galerkin method